WO2025199869A1

WO2025199869A1 - Methods and apparatus of measurement overhead and measurement gap reduction with ai/ml prediction

Info

Publication number: WO2025199869A1
Application number: PCT/CN2024/084423
Authority: WO
Inventors: Ta-Yuan LIU; Per Johan Mikael Johansson; Yao PENG; Yuanyuan Zhang
Original assignee: MediaTek Inc
Current assignee: MediaTek Inc
Priority date: 2024-03-28
Filing date: 2024-03-28
Publication date: 2025-10-02
Anticipated expiration: 2026-09-28
Also published as: WO2025201451A1

Abstract

Apparatus and methods are provided for UE to perform measurement and/or prediction of multiple measurement objects (MO) subject to restrictions that inhibit the UE from measuring or detecting all MOs. The UE selects MO for measurement and/or prediction from different subsets, characterized by selection and prediction intended to be done by UE internal intelligence based on collected data. The MO measurement prediction and selection can be used by UE to reduce the measurement overhead and the need of measurement gap (MG). A general procedure is provided to perform the measurement overhead reduction and MG by AI measurement reduction, which contains the steps of configuration, measurement, prediction, and report.

Description

METHODS AND APPARATUS OF MEASUREMENT OVERHEAD AND MEASUREMENT GAP REDUCTION WITH AI/ML PREDICTION

FIELD

The present disclosure relates generally to communication systems, and more particularly, the method and apparatus of measurement overhead and measurement gap reduction by measurement prediction.

BACKGROUND

In wireless communication systems, e.g., NR, LTE, 3gpp network, etc, the UE constantly measures the multiple measurement objects (MOs) to measure and detect the beam and cell (and/or their quality) for multiple uses, e.g., handover, RRM. The measurement of MO is one of the key overheads at the UE. Furthermore, when the target MO is not located in the same frequency layer as the current operating setting, the measurement is needed and UE can only measure the MO within the measurement occasions due to the RF limitation.

One solution to reduce the overhead is to predict the MO measurement result based on the previous measurement sample. Artificial Intelligence (AI) and Machine Learning (ML) is one of the potential approaches to predict the measurement, which have permeated a wide spectrum of industries, ushering in substantial productivity enhancements. In the realm of mobile communications systems, these technologies are orchestrating transformative shifts. Mobile devices are progressively supplanting conventional algorithms with AI-ML models.

Once the measurement of certain MO can be predicted without actual measurement, the UE can reduce the measurement overhead. Additionally, the UE is possible to request a measurement gap configuration with a longer period. If the network, e.g., gNB, acknowledges the request, the longer measurement gap can provide a high transmission capability, i.e., part of the original measurement gap duration can be used for either DL or UL transmission.

In this invention, apparatus and mechanisms are sought to provide the frame work and the corresponding procedure to reduce the measurement overhead and MG by measurement prediction.

SUMMARY BRIEF DESCRIPTION OF THE DRAWINGS DETAILED DESCRIPTION

In this invention, methods and apparatus are provided for UE to perform measurement and/or prediction of multiple measurement objects (MO) subject to restrictions that inhibit the UE from measuring or detecting all MOs at full performance, where the UE selects subsets of MOs for measurement and/or prediction, characterized by selection and prediction intended to be done by UE internal intelligence based on collected data

MOs have reference pilot transmissions for measurements and are identified by one or more elements including frequency, e.g. ARFCN, ERFCN, timing information, type of pilot/symbols, bandwidth, and pattern information. One typical embodiment that includes MO measurement is NR SSB and/or CSI-RS RRM measurement of neighbor cells. In another embodiment, beams are detected or measured.

The restriction that inhibits the UE from measuring or detecting all MOs comes from the physical and/or RF limitation that UE needs to reconfigure/switch its radio hardware configuration specifically for different MOs, for example, to switch frequency to MOs on a frequency that is different from the current UE operating frequency. In one embodiment, UE has a measurement gap configuration from the network. Due to the physical and/or hardware limitation, the UE measures and/or detects those cells or beams, that can not be observed by the current operating configuration, within the measurement gap.

Figure 1 provides an exemplary illustration of the NR SSB measurement. UE performs MO measurements of three different frequency layers. With the RF limitation, the UE measures different MOs (or different Rx beams) in turn.

In this invention, a method is provided for UE to reduce the overhead of MO measurement or the need for measurement gap by MO prediction and/or MO selection. In one embodiment, the prediction and selection is performed by AI-ML approaches, e.g., a data-driven approach based on the pre-training neural network model.

In one embodiment, the UE uses previous measurement samples of MOs to predict future measurement of MO or measurement of other MOs. Based on the prediction, the UE does not need to measure all MO and thus the measurement overhead and/or the need of measurement gap can be reduced.

Figure 2 provides an exemplary illustration that the UE uses previous measurement samples of an MO, to predict future measurement results of this MO. With the help of prediction, UE does not need to measure MO for each cycle (e.g., the red part in Figure 2) , as a result, the measurement overhead can be reduced and/or the measurement gap (MG) cycle can be enlarged which reduces the need of MG. In one embodiment, the UE applies AI temporal domain prediction based on the previous measurement samples to predict the target MO.

In another embodiment, the UE uses measurement samples of other MOs, to predict measurement results of the target MO. This includes the use of AI spatial domain prediction, where the AI predicts the result of a certain MO based on the measurement of the same and/or other MOs with spatial correlation, for example, predicts the measurement of MO with certain Tx (or Rx) beam by the measurement of other Tx (or Rx) beams. This also includes the use of AI inter-frequency prediction, where the AI predicts the result of certain MO based on the measurement of other MOs with different frequencies. The correlation between MOs with different frequencies could come from the geographical dependence, for example, two different frequency carriers are transmitted from co-located gNB.

In one embodiment, a joint temporal and frequency domain prediction is applied for UE to make the MO prediction, which is illustrated in Figure 3.

Figure 4 provides an exemplary illustration of the use of MO prediction, where the measurement predictions can be directly used as the real measurement of certain MO and/or can be used to prioritize/select MOs that need to be measured later.

In one embodiment, the UE performs MO selection by predicting measurement result for multiple MO and prioritizes the MO (s) with the highest predicted measurement value (s) . The measurement value can be the maximum RSRP of cells measured in the MO, the average RSRP of cells measured in the MO, or other values that can represent the quality of the cells in the MO.

In one embodiment, the UE excludes/deprioritizes MOs for which the predicted measurement values are below a criterion /a pre-defined or configured threshold. The measurement value can be the maximum RSRP of cells measured in the MO, the average RSRP of cells measured in the MO, or other values that can represent the quality of the cells in the MO.

In one embodiment, the UE excludes/deprioritizes those MOs whose predictions are expected to have high accuracy, e.g., for those MOs that have a high correlation with previous measurement samples.

The UE additionally uses correlation information between the predicted MO and other MOs, for example, to what extent both MOs experience similar or different penetration loss. In one embodiment, the correlation between two MOs is low when one is an indoor base station and the other is an outdoor base station. In another embodiment, the correlation is high when the reference pilot transmissions for the MOs are transmitted from the same transmission point

In one embodiment, the UE uses the correlation information to determine to what extent it can rely on prediction and to what extent it needs to perform actual measurements.

In one embodiment, the correlation information is provided by the NW, e.g., gNB, to the UE by signaling. The correlation information can be learned separately, e.g., per cell, and/or learned with multiple cells.

In one embodiment, the indoor/outdoor indicator is included in the correlation information. The correlation between two MOs that belong to different values, e.g., one is indoor and the other is outdoor, is assumed to be zero (or low) .

In another embodiment, as shown in Figure 5, the typical ToS time for cells of a certain MO is considered as part of correlation information. It can be used to determine how often the UE needs to actually measure to consolidate a certain prediction. For example, the UE can perform lower frequent measurements for the MO with the longer ToS and vice versa.

The location-related info, e.g., TA for both UE and gNB side, RTT info, …etc. is provided to derive the MO correlation. In one embodiment, the location info is derived by on UE based eCID positioning with gNB providing additional TA info to the UE by RRC.

In one embodiment, the UE uses geographical area information to prioritize such MOs for selection, for example, geographical areas where a MO is predicted to be above a criterion /a pre-defined or configured threshold is stored and used. In another embodiment, geographical information is part of the input to AI progress, e.g. a neural network.

The geographical information includes not only popular geo-coordinate formats, but also cellular/wireless specific formats: cell ID, e.g. complemented with RTT information etc.

Figure 6 illustrates the exemplary general procedure that UE reduces the measurement overhead and/or the need of measurement gap by measurement prediction, which comprises the steps of configuration, measurement, prediction, and report.

The configuration step refers to the procedure to transmit the information of two subsets of MOs. One contains the information related to those MOs that should be measured, the other contains the information related to those MO that could be predicted.

The measurement step refers to the procedure in which UE performs MO measurement based on the configuration.

The prediction step refers to the procedure in which UE performs MO measurement prediction based on the configuration and previous measurement samples of MOs.

The report step refers to the procedure in which UE triggers the measurement report to report the MO measurement to the gNB. It also includes the procedure in which UE sends the request to gNB to indicate the capability that UE can reduce the need for the measurement gap by measurement prediction.

Figure 7 illustrates the exemplary procedure in which UE performs the measurement prediction based on the configuration and previous measurement to reduce the measurement overhead and to request for the reduction of the need for measurement gap.

In one embodiment, the UE first transits the UE capability report to inform the capability of UE measurement prediction. The capability report contains one or more elements including AI capability, temporal domain prediction capability, spatial domain prediction capability, inter-freq prediction capability, available measurement gap cycle, measurement reduction ratio, etc. The measurement reduction ratio indicator the possible reduce of measurement compared with those UE does not have measurement prediction capacity.

The UE received the measurement configuration from the gNB, where the measurement configuration contains the configuration of two different sets of MOs. The first set, e.g., the measured set, indicates the MO needed to be measured. The configuration contains the measurement occasion and measurement information for each MO in the set, e.g., the reference/pilot signal, e.g., SSB, the smtc, the time and freq resource that the reference/pilot signal is transmitted, …etc. The second set, e.g., the predicted set, indicates those MO that UE can use prediction instead of actual measurements. The configuration of the second set (predicted set) includes the prediction occasion, the type of virtual reference/pilot signal, e.g., SSB, CSI-RS, etc.

The measurement configuration provides the measurement gap configuration.

The association between the measurement occasions, e.g., the occasions for the first set, and the prediction occasions, e.g., the occasion for the second set, is configured.

In one embodiment, the first set and the second set of measurement objectives can be the same, for example, for temporal domain prediction use case.

In another embodiment, the association between the measurement occasions in the first set of measurement objects and the prediction occasions in the second set of measurement objects is configured, for example, for the temporal domain combine with frequency domain prediction use case.

In another embodiment, the association between the first set of measurement objects and the second set of measurement objects for measurement and prediction is configured, for example, for frequency domain prediction

For each measurement occasion and prediction occasions, one or multiple beams in spatial domain are considered.

After receiving the measurement configuration, the UE performs measurement on the measurement occasions configured for the first set based on the configuration of the first set, the association between the second set, and the measurement gap configuration.

In one embodiment, the UE selects the measurement target based on the measurement prediction. For each occurrence of the measurement occasions of the first set, UE selects the top measurement object to measure which meets the pre-defined criterion based on the predicted measurement results. In one embodiment, the measurement object that has the best-predicted measurement result in the measurement occasion is selected. In another embodiment, The measurement object that has the best-predicted measurement results in the next consecutive measurement occasions is selected.

UE preforms measurement prediction on the prediction occasions for the second set of measurement objects.

In one embodiment, the UE performs temporal domain prediction. In another embodiment, the UE performs frequency domain prediction. In another embodiment, the UE performs spatial domain prediction. In another embodiment, the UE performs prediction with the combination of temporal domain, spatial domain, and frequency domain.

UE evaluates the triggering events based on measured results, predicted results, or combination of them. In one embodiment, the triggering events contain the measurement report event, e.g., A1～A6, B1, B2 events.

UE send reports with measured results, predicted results, or combination of them.

Additionally, UE figures out the proper measurement gap and send the measurement gap configuration to the network. The measurement gap configuration can consider the periodicity, offset and duration. UE can apply the proper measurement gap no matter network ACKed it or not as long as the measurement gap has less measurement opportunities than the one configured by the network. Alternatively, UE sends the UE capability on the MG patterns it can support (based on the AI/ML capability) and network provide the MG configuration to UE.

Claims

A method for UE to perform measurement and/or prediction of multiple measurement objects (MO) subject to restrictions that inhibit the UE from measuring or detecting all MOs at full performance, where the UE selects subsets of MOs for measurement and/or prediction, characterized by selection and prediction intended to be done by UE internal intelligence based on collected data.
The method of claim 1, wherein MOs have reference pilot transmissions for measurements and are identified by one or more elements including frequency, e.g. ARFCN, ERFCN, timing information, type of pilot/symbols, bandwidth, and pattern information.
The method of claim 2, wherein the MO is measured for NR SSB or CSI-RS RRM measurements.
The method of claim 2, wherein the MO is measured for beam detection and/or beam measurement.
The method of claim 1, wherein the restriction comes from the physical and/or RF limitation that UE needs to reconfigure/switch its radio hardware configuration specifically for different MOs.
The method of claim 5, UE receives a measurement gap configuration from the network, the UE switches frequency to MOs on a frequency that is different from the current operating setting, during the measurement gap occasions, to measure and/or detect those cells or beams in the MOs.
A method for UE to reduce the MO measurement overhead and/or the need for measurement gap by MO prediction and/or MO selection.
The method of claim 1 and 7, wherein the prediction and selection are performed by AI-ML approaches, e.g., a data-driven approach based on the pre-training neural network model, and/or non-AI approaches based on the previous MO measurement samples.
The method of claim 8, wherein the MO prediction is performed by temporal domain prediction, where the UE uses previous measurement samples of an MO, to predict future measurement results of this MO.
The method of claim 8, wherein the MO prediction is performed by spatial domain prediction, where the UE predicts the result of a certain MO based on the measurement of the same and/or other MOs with spatial correlation.
The method of claim 10, wherein the UE predicts the measurement of MO with certain Tx (or Rx) beams by the measurement of other Tx (or Rx) beams of the same MO.
The method of claim 8, wherein the MO prediction is performed by inter-freq prediction, the UE predicts the result of a certain MO by the measurement of the other MOs with different frequencies based on geographical dependence, for example, the pilot/reference signal for those MOs is transmitted from co-located gNB.
The method of claim 8, wherein the MO prediction is performed by considering joint temporal, spatial, and frequency domain prediction.
The method of claim 8, wherein the measurement predictions are directly used as the real measurement of certain MO.
The method of claim 8, wherein measurement predictions are used to prioritize/select MOs that need to be measured.
The method of claim 15, wherein the UE performs MO selection by predicting measurement results for multiple MO and prioritizes the MO (s) with the highest predicted measurement value (s) , the measurement value can be the maximum RSRP of cells measured in the MO, the average RSRP of cells measured in the MO, or other values that can represent the quality of the cells in the MO.
The method of claim 15, wherein the UE excludes/deprioritizes MOs for which the predicted measurement values are below a criterion /a pre-defined or configured threshold, the measurement value can be the maximum RSRP of cells measured in the MO, the average RSRP of cells measured in the MO, or other values that can represent the quality of the cells in the MO.
The method of claim 8, further comprising UE uses correlation information between the predicted MO and other measured MOs for prediction and selection.
The method of claim 18, wherein correlation information is used to determine which MO measurement can rely on prediction and which is needed to perform actual measurements.
The method of claim 18, wherein the correlation information is provided by the NW, e.g., gNB, to the UE by signaling, the correlation information can be learned separately, e.g., per cell, and/or learned with multiple cells.
The method of claim 18, wherein the correlation information contains the indoor/outdoor indicator, the correlation between two MOs that belong to different values, e.g., one is indoor and the other is outdoor, is assumed to be zero (or low) .
The method of claim 18, wherein the typical ToS time for cells of a certain MO is considered as part of correlation information, the UE performs lower frequent measurements for the MO with the longer ToS and vice versa.
The method of claim 18, further comprising UE receives the location-related info, e.g., TA for both UE and gNB side, RTT info, …etc. for deriving the MO correlation.
The method of claim 23, wherein the location info is derived by UE based eCID positioning with gNB providing additional TA info to the UE by RRC.
The method of claim 8, further comprising the UE uses geographical area information to prioritize MOs for selection.
The method of claim 25, wherein the geographical areas information contains the information that a set of MO is predicted to be above a criterion /a pre-defined or configured threshold.
The method of claim 25, wherein the geographical information is part of the input to AI progress, e.g. a neural network.
The method of claim 25, wherein the geographical information includes not only popular geo-coordinate formats, but also cellular/wireless specific formats: cell ID, e.g. complemented with RTT information etc.
The method of claim 7, further comprising the general procedure that contains the step of configuration, measurement, prediction, and report.
The method of claim 29, further comprising the UE first transits the UE capability report to inform the capability of UE measurement prediction, the capability report contains one or more elements including AI capability, temporal domain prediction capability, spatial domain prediction capability, inter-freq prediction capability, available measurement gap cycle, measurement reduction ratio, etc; the measurement reduction ratio indicator the possible reduction of measurement compared with those UE does not have measurement prediction capacity.
The method of claim 29, wherein the UE received the measurement configuration from the gNB, where the measurement configuration contains the configuration of two different sets of MOs; The first set, e.g., the measured set, indicates the MO needed to be measured; The second set, e.g., the predicted set, indicates those MO that UE can use prediction instead of actual measurements.
The method of claim 31, wherein the configuration contains the measurement occasion and measurement information for each MO in the set, which contains one or more elements including the type of reference/pilot signal, e.g., SSB, CSI-RS, the smtc, the time and freq resource that the reference/pilot signal is transmitted, measurement occasion, prediction occasion, and the type of virtual reference/pilot signal, e.g., the type of signal for the predicted MO, etc.
The method of claim 31, wherein the measurement configuration contains measurement gap configuration.
The method of claim 31, wherein the association between the measurement occasions, e.g., the occasions for the first set, and the prediction occasions, e.g., the occasion for the second set, is configured.
The method of claim 31, wherein, the first set and the second set of measurement objects are the same.
The method of claim 31, wherein for each measurement occasion and prediction occasion, one or multiple beams in spatial domain are considered.
The method of claim 29, wherein the UE performs measurement on the measurement occasions configured for the first set based on the configuration of the first set, the association between the second set, and the measurement gap configuration.
The method of claim 29, further comprising the UE selects the measurement target based on the measurement prediction; For each occurrence of the measurement occasions of the first set, UE selects the measurement object that meets the pre-defined criterion based on the predicted measurement results.
The method of claim 38, wherein the UE selects the measurement object that has the best-predicted measurement result in the measurement occasion.
The method of claim 38, wherein the UE selects measurement object that has the best-predicted measurement results in the next consecutive measurement occasions is selected.
The method of claim 29, wherein UE performs measurement prediction on the prediction occasions for the second set of measurement objects.
The method of claim 29, wherein UE evaluates the triggering events based on measured results, predicted results, or a combination of them.
The method of claim 42, wherein the triggering events include the measurement report event, e.g., A1～A6, B1, B2 events.
The method of claim 29, wherein UE send reports with measured results, predicted results, or a combination of them.
The method of claim 29, wherein UE figures out the proper measurement gap and send the measurement gap configuration to the network; The measurement gap configuration contains the periodicity, offset and duration.
The method of claim 29, wherein UE apply the proper measurement gap no matter network ACKed it or not as long as the measurement gap has less measurement opportunities than the one configured by the network; Alternatively, UE sends the UE capability on the MG patterns it can support (based on the AI/ML capability) and network provide the MG configuration to UE.